基于统计学模型的沟道侵蚀敏感性评估的研究进展与展望

侵蚀沟的发生通常会造成区域严重的环境破坏和经济损失。开展沟道侵蚀敏感性评估有助于相关部门采取合理有效的措施抑制并降低沟蚀的发生，对改善区域环境、维持粮食生产能力，保障经济健康发展具有重要意义。该文从沟道侵蚀敏感性评估流程、数据基础以及统计学模型的应用3个方面，系统地介绍了该方向的研究进程，列举了各分类下统计学模型的优缺点，分析了该方向的相关研究进展，对比了不同研究之间的应用条件，并指出，未来研究还需实现迁移学习和时间序列模型在沟道侵蚀敏感性评估中的应用，以及深度学习模型与侵蚀沟发育物理机制相互融合，从而为改进和加强统计学模型与沟道侵蚀敏感性评估之间交叉应用，推动土壤侵蚀学科发展，为区域发展规划奠定理论与技术基础。

Research advances and prospects on gully erosion susceptibility assessment based on statistical modeling

Gully erosion can usually cause the serious environmental damage and economic losses. Susceptibility assessment and spatial distribution of gully erosion can provide a strong references for the targeted measures, decision-making on environmental protection, geologic disaster prevention, water resource management, and infrastructure planning. It is of great significance to better understand and manage the gully system, and then reduce potential risks and losses. The regional environment can be improved to maintain the regional food production in sustainable economic development. Much effort has focused on the prediction models of soil erosion over the past few decades. The amount of erosion on slopes has been predicted from both empirical statistical and physical genesis models. However, few models have been developed to predict the development of erosion gullies. Gully erosion is more complicated than slope erosion. Furthermore, the modeling has only focused on the shallow gully erosion. Only a few studies have been conducted on the topographic, critical and empirical model of gully morphologic features, the gully erosion prediction, and the landscape evolution. It is still lacking on general extrapolation of these models, even the susceptibility of gully erosion. Multiple factors can also be considered, such as the topography, geomorphology, soil, climate, vegetation and anthropogenic factors. A comprehensive and quantitative analysis of the influencing factors can be expected with the remote sensing satellites, resource and environmental surveys. A large amount of gully-erosion data has been accumulated to statistically modeling in this field. The susceptibility assessment of gully erosion aims to calculate the importance of the influencing factors on the occurrence of gully erosion and the prediction performance between different algorithms. It is still lacking on the innovative and promotional research in this field. In this review, the flow of research in this direction was systematically introduced to summarize the strengths and weaknesses of the statistical models under various classifications. The advance in this direction was proposed to compare the commonalities and differences in the application conditions from three aspects: the process of gully erosion, the data construction, and the application of statistical model. At the same time, future research needs to realize the application of transfer learning and time series models in the susceptibility assessment on gully erosion. The deep learning and physical mechanisms can be integrated to clarify the erosion gully development, in order to improve and strengthen the cross-disciplinary application of statistical models in soil erosion. The finding can lay the theoretical and technical foundation for regional development planning.